Description

You’ve decided to open a small robot-run cafe in Los Angeles. The project is promising but expensive, so you and your partners decide to try to attract investors. They’re interested in the current market conditions—will you be able to maintain your success when the novelty of robot waiters wears off? You’re an analytics guru, so your partners have asked you to prepare some market research. You have open-source data on restaurants in LA.

Link to presentation https://drive.google.com/file/d/1TVUoAm0YTSeMQf9cgDr5X0IuSHaHDvxh/view?usp=sharing

!pip install usaddress

Collecting usaddress
  Downloading usaddress-0.5.10-py2.py3-none-any.whl (63 kB)
Requirement already satisfied: future>=0.14 in c:\users\edeng\anaconda3\lib\site-packages (from usaddress) (0.18.2)
Collecting python-crfsuite>=0.7
  Downloading python_crfsuite-0.9.8-cp39-cp39-win_amd64.whl (158 kB)
Collecting probableparsing
  Downloading probableparsing-0.0.1-py2.py3-none-any.whl (3.1 kB)
Installing collected packages: python-crfsuite, probableparsing, usaddress
Successfully installed probableparsing-0.0.1 python-crfsuite-0.9.8 usaddress-0.5.10

Preprocessing data

import pandas as pd
from plotly import graph_objects as go 
import plotly.express as px 
import usaddress
import numpy as np
import seaborn as sns
from matplotlib import pyplot as plt

rest_data=pd.read_csv('/Users/edeng/Downloads/rest_data_us.csv')
rest_data.info()
#show nan values
print(rest_data[rest_data.isna().any(axis=1)])
#we have 3 rows with nan values in the chain column. 
#As it is impossible to fill this up, I will drop them
rest_data=rest_data.dropna()
#checking duplicates
rest_data[rest_data.duplicated()]

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 9651 entries, 0 to 9650
Data columns (total 6 columns):
 #   Column       Non-Null Count  Dtype 
---  ------       --------------  ----- 
 0   id           9651 non-null   int64 
 1   object_name  9651 non-null   object
 2   address      9651 non-null   object
 3   chain        9648 non-null   object
 4   object_type  9651 non-null   object
 5   number       9651 non-null   int64 
dtypes: int64(2), object(4)
memory usage: 452.5+ KB
         id              object_name               address chain object_type  \
7408  19194  TAQUERIA LOS 3 CARNALES  5000 E WHITTIER BLVD   NaN  Restaurant   
7523  19309     JAMMIN JIMMY'S PIZZA   1641 FIRESTONE BLVD   NaN       Pizza   
8648  20434    THE LEXINGTON THEATER          129 E 3RD ST   NaN  Restaurant   

      number  
7408      14  
7523       1  
8648      35  

	id	object_name	address	chain	object_type	number

Investigate the proportions of the various types of establishments. Plot a graph.

rest_prop=rest_data.groupby('object_type')['id'].count().reset_index()
rest_prop
fig = go.Figure(data=[go.Pie(labels=rest_prop['object_type'], values=rest_prop['id'], title='Establishment Proportions')])

fig.show() 

The majority of our data contain info about Restaurant types, while the least of our data contain info about Bakery types.

Investigate the proportions of chain and nonchain establishments. Plot a graph.

chain_prop=rest_data.groupby('chain')['id'].count().reset_index()
chain_prop
fig = go.Figure(data=[go.Pie(labels=chain_prop['chain'], values=chain_prop['id'], title='Is it a chain?')])
fig.show()

62% of the data is non chain establishments, while 38% are chains.

Which type of establishment is typically a chain?

typical_chain=rest_data.groupby(['chain', 'object_type'])['id'].count().reset_index()
typical_chain.columns=['chain', 'Establishment', 'establishment_count']
fig = px.bar(typical_chain, x='Establishment', y='establishment_count', color='chain', title='Which type of establishment is typically a chain?')
fig.update_xaxes(tickangle=45)
fig.show() 
v=typical_chain.Establishment.value_counts()
typical_chain[~typical_chain.Establishment.isin(v.index[v.gt(1)])]

	chain	Establishment	establishment_count
5	True	Bakery	283

Bakery is the establishment type which is typically a chain, as it is the only establishment which has no instances where it isn't a chain.

What characterizes chains: many establishments with a small number of seats or a few establishments with a lot of seats?

chain_char=rest_data[rest_data['chain']==True]
chain_char=chain_char.groupby(['object_name', 'object_type'])['number'].agg(['sum','count']).reset_index()
chain_char.columns=['object_name', 'establishment', 'number_of_seats', 'number_of_establishments']
print('The average seats per establishment is:', chain_char['number_of_seats'].sum()/chain_char['number_of_establishments'].sum())
print('The mean number of seats per chain:', chain_char['number_of_seats'].mean())
print('The mean number of establishments per chain:', chain_char['number_of_establishments'].mean())
chain_char=chain_char.groupby('establishment')[['number_of_seats', 'number_of_establishments']].sum().reset_index()
fig = px.bar(chain_char, x='establishment', y='number_of_seats', color='number_of_establishments', title='Number of seats to number of establishments in chains')
fig.update_xaxes(tickangle=45)
fig.show() 
chain_char['seats_average']=chain_char['number_of_seats']/chain_char['number_of_establishments']
fig = px.bar(chain_char, x='establishment', y='seats_average', title='Seats ratio per chain establishment')
fig.update_xaxes(tickangle=45)
fig.show() 

The average seats per establishment is: 39.69423286180631
The mean number of seats per chain: 53.35137111517368
The mean number of establishments per chain: 1.3440585009140769

There's an average of 40 seats for each establishment which indicate that there are few establishment with a lot of seats as a characterization of chains. We can also see that there are not a lot of establishments per chain, while average seats per chain is high at 53 seats.

Determine the average number of seats for each type of restaurant. On average, which type of restaurant has the greatest number of seats? Plot graphs.

avg_seats=rest_data.groupby('object_type')['number'].mean().reset_index().sort_values('number', ascending=False)
fig = px.bar(avg_seats, x='object_type', y='number', labels={
                     "object_type": "Restaurant Type",
                     "number": "Average Number of Seats"
                 }, title='Average seats by restaurant type')
fig.show()

Restaurants has the greatest number of seats.

Put the data on street names from the address column in a separate column.

def cleaning(raw):
    raw_address=usaddress.parse(raw)
    PostType= [item for item in raw_address if 'StreetNamePostType' in item]
    streetName=[item for item in raw_address if 'StreetName' in item]
    streetName=dict(streetName)
    if PostType:
        if (len(streetName)) > 1:
            streetName=' '.join(map(''.join, streetName.keys()))
            return streetName + ' ' + PostType[0][0]
        else:
            try:
                streetName=sorted(streetName.keys(), reverse=True)
                return streetName[0] + ' ' + PostType[0][0]
            except:
                return np.nan
    else:
        if (len(streetName)) > 1:
            streetName=' '.join(map(''.join, streetName.keys()))
            return streetName
        else:
            try:
                streetName=sorted(streetName.keys(), reverse=True)
                return streetName[0]
            except:
                return np.nan

rest_data['street_name']=rest_data.address.apply(cleaning)

rest_data

	id	object_name	address	chain	object_type	number	street_name
0	11786	HABITAT COFFEE SHOP	3708 N EAGLE ROCK BLVD	False	Cafe	26	EAGLE ROCK BLVD
1	11787	REILLY'S	100 WORLD WAY # 120	False	Restaurant	9	WORLD WAY
2	11788	STREET CHURROS	6801 HOLLYWOOD BLVD # 253	False	Fast Food	20	HOLLYWOOD BLVD
3	11789	TRINITI ECHO PARK	1814 W SUNSET BLVD	False	Restaurant	22	SUNSET BLVD
4	11790	POLLEN	2100 ECHO PARK AVE	False	Restaurant	20	ECHO PARK AVE
...	...	...	...	...	...	...	...
9646	21432	HALL OF JUSTICE	217 W TEMPLE AVE	False	Restaurant	122	TEMPLE AVE
9647	21433	FIN-MELROSE	5750 MELROSE AVE	False	Restaurant	93	MELROSE AVE
9648	21434	JUICY WINGZ	6741 HOLLYWOOD BLVD	True	Fast Food	15	HOLLYWOOD BLVD
9649	21435	MEDIDATE COFFEE	548 S SPRING ST STE 100	False	Cafe	6	SPRING ST
9650	21436	CAFE SPROUTS	1300 S SAN PEDRO ST STE 111	True	Restaurant	19	SAN PEDRO ST

9648 rows × 7 columns

#checking nan street names
print(rest_data[rest_data['street_name'].isna()].reset_index())
# filling the nan values
rest_data[(rest_data['address'].str.contains('1033 1/2 LOS ANGELES ST'))]=rest_data[(rest_data['address'].str.contains('1033 1/2 LOS ANGELES ST'))].apply(lambda x: x.fillna('LOS ANGELES ST'))
rest_data[(rest_data['address'].str.contains('OLVERA ST'))]=rest_data[(rest_data['address'].str.contains('OLVERA ST'))].apply(lambda x: x.fillna('OLVERA ST'))
rest_data[rest_data['street_name'].isna()]

   index     id           object_name                  address  chain  \
0   1888  13674      MAGDALENO DEBBIE            OLVERA ST E17  False   
1   3719  15505            LAS ANITAS             OLVERA ST 26  False   
2   7003  18789           D.K. DONUTS  1033 1/2 LOS ANGELES ST  False   
3   7281  19067         CIELITO LINDO             OLVERA ST 23  False   
4   7319  19105        JUANITA'S CAFE             OLVERA ST 20  False   
5   7321  19107    KITTY'S SNOW CONES              OLVERA ST 2  False   
6   7438  19224         LA GOLONDRINA            OLVERA ST W17  False   
7   7741  19527  RODOLFO'S RESTAURANT              OLVERA ST 5   True   

  object_type  number street_name  
0  Restaurant      15         NaN  
1  Restaurant      38         NaN  
2   Fast Food       8         NaN  
3  Restaurant       2         NaN  
4  Restaurant      12         NaN  
5  Restaurant      11         NaN  
6  Restaurant     204         NaN  
7  Restaurant       4         NaN  

	id	object_name	address	chain	object_type	number	street_name

Plot a graph of the top ten streets by number of restaurants.

top_10=rest_data.groupby('street_name')['id'].count().reset_index().sort_values('id', ascending=False).head(10)
top_10

	street_name	id
475	SUNSET BLVD	403
537	WILSHIRE BLVD	397
411	PICO BLVD	370
523	WESTERN AVE	367
211	FIGUEROA ST	332
393	OLYMPIC BLVD	309
502	VERMONT AVE	286
442	SANTA MONICA BLVD	264
27	3RD ST	260
261	HOLLYWOOD BLVD	253

fig = px.bar(top_10, x='street_name', y='id', labels={
                     "street_name": "Street Name",
                     "id": "Number of Restaurants"
                 }, title='Number of restaurants by street')
fig.show()

Sunset BLVD has the highest number of restaurants at close to 400, followed by Wilshire BLVD and Pico BLVD, with 397 and 370 restaurants respectively. Number 10 on the list is Hollywood BLVD with 253 restaurants.

Find the number of streets that only have one restaurant.

only_one=rest_data.groupby('street_name')['id'].count().reset_index()
only_one=only_one[only_one['id']==1].count()
only_one

street_name    250
id             250
dtype: int64

There are 250 street with only 1 restaurant.

For streets with a lot of restaurants, look at the distribution of the number of seats. What trends can you see?

top_rest=rest_data.groupby('street_name')['number'].agg(['sum', 'count']).reset_index().sort_values('count', ascending=False).head(10)
top_rest.columns=['street_name', 'seats_number', 'restaurants_number']
top_rest
fig = px.scatter(top_rest, x="restaurants_number", y="seats_number", color='street_name', size='restaurants_number')
fig.show()

sns.pairplot(top_rest);

From the above we can clearly see a positive correlation in the number of the restaurant with the number of seats. The higher the number of restaurants in the street, there is an increase in the number of seats.

#box plot 
top_rest.set_index('street_name', inplace=True)

#creating a function that will show the names of the streets on the outliers
def make_labels(ax, boxplot):
    fly = boxplot['fliers'][0]
    med = boxplot['medians'][0]
    # The x position of the median line
    xpos = med.get_xdata()

    # Lets make the text have a horizontal offset which is some 
    # fraction of the width of the box
    xoff = 0.10 * (xpos[1] - xpos[0])

    # The x position of the labels
    xlabel = xpos[1] + xoff

    # The median is the y-position of the median line
    median = med.get_ydata()[1]
    
    for flier in fly.get_ydata():
        ax.text(1 + xoff, flier,
                '{}'.format(top_rest.index.where(top_rest.seats_number==flier).dropna()[0]), va='center')
        
red_diamond = dict(markerfacecolor='r', marker='D')
fig, ax = plt.subplots()
ax.set_title('Seats number boxplot')

# Create the boxplot and store the resulting python dictionary
my_boxes = ax.boxplot(top_rest['seats_number'], flierprops=red_diamond)

# Call the function to make labels
make_labels(ax, my_boxes)
plt.xlabel('Seats Number')
plt.show()

We can see the street names where the number of seats are outliers in our data.

In conclusion, I inspected the data and viewed 3 Nan values in the chain field, which I removed as they were irrelevant to my task. There were no duplicates.

I viewed that 75% of the data is comprised of restaurants and 11% of fast food establishments, in addition 61% of the data is related to chains.

I found that bakeries in the data is only appearing as a chain. In addition I saw that chains tends to have fewer establishments with many seats, with restaurants having the highest average number of seats.

Lastly I inspected the 10 top streets in terms of establismnets count, and also inspected that 250 streets have 1 restaurant, and found that there is a positive correlation between the number of establishments on the street, and the number of seats.

Cafe establishments have 25 seats on average and consist of about 50% chains.

Per the above conclusion it may be recommended to open an average size cafe (25 seats) in a location which is not very packed with restaurants, but also don't have only 1. I recommed looking to open on streets which appear on the top 10 streets but also do not appear as an 3rd quarter outlier on the boxplot above, as it may indicate of overcrowding.